Notes on operating inference.
Launch notes, strategy explainers, and deep-dives — written by the people who run the gateway.
LaunchArchitecture
One endpoint, every provider: why we built Synapass
Inference stacks accrete instead of being chosen. Synapass moves routing, failover, and cost control into one operated layer.
2026-09-18 6 min readRead
RoutingGuide
Routing strategies, explained: when to use each one
priority, weighted, lowest_cost, lowest_latency, highest_quality — what each strategy optimizes, and how fallback chains interact with them.
2026-09-25 8 min readRead
CachingGuide
Caching LLM responses without the footguns
Exact, prefix, and semantic tiers with tenant isolation and scoped invalidation — plus why caching is off by default.
2026-10-01 7 min readRead